Last updated: 5 September 2026
RadioCheck stands on other people's work. This page lists it, says what licence each part is used under, and tells you where to get the source. If we have got something wrong here, write to hi@radiocheckapp.com and we will fix it.
Two pieces of software inside RadioCheck are not libraries linked into the app. They are standalone programs that RadioCheck starts and talks to, the way any app might call a command line tool. They keep their own licences, and you can replace them.
Alan's speech is produced by Piper, which embeds eSpeak NG. Both are licensed under the GNU General Public License, version 3 or later.
Because the GPL requires that any program linking that code also be GPL, RadioCheck does
not link it. Piper lives in a separate program called piper-worker, which is
itself GPLv3 and whose complete source is distributed with the app (inside the application
bundle, and in the RadioCheck repository under tools/piper_worker/). RadioCheck
communicates with it only through its standard input and output.
You may obtain, modify and redistribute piper-worker under the terms of the
GPLv3. Write to hi@radiocheckapp.com if you want
the source separately and we will send it.
FFmpeg is used to decode and encode audio. RadioCheck ships a build made specifically for it, compiled without the GPL-only components, so it is used under the GNU Lesser General Public License, version 2.1 or later. It includes LAME (LGPL) for MP3 encoding and libopus (BSD).
The exact build recipe is published with RadioCheck
(tools/build_ffmpeg_lgpl.sh), and the configuration used is recorded inside the
app bundle. The LGPL lets you replace these libraries with your own build; the licence text
ships alongside them.
The voice models are not included in the download. RadioCheck fetches the one for your language during setup, from the official Piper voices repository, pinned to a fixed revision.
| Voice | Language | Source |
|---|---|---|
| en_GB-alan-medium | English | Recorded by Alan Pope, and used here with his personal permission, kindly given in September 2026. Alan the race engineer is named after this voice, and (as it turns out) after a real person too. Thank you, Al. |
| es_ES-davefx-medium | Spanish | Recorded by the contributor known as davefx and released under CC0 (public domain) through the OHF Voice datasets. Thank you as well, whoever you are. |
Speech recognition uses Whisper models (MIT), also downloaded on demand rather than bundled.
All permissively licensed, all used unmodified.
| Component | Licence | Used for |
|---|---|---|
| discord.py · discord-ext-voice-recv | MIT | Alan's voice in a Discord channel, and hearing you |
| PyNaCl | Apache 2.0 | Discord voice encryption |
| opuslib | BSD 3-Clause | Opus audio for Discord |
| faster-whisper · CTranslate2 | MIT | Speech recognition on your Mac |
| onnxruntime | MIT | Running the voice and speech models |
| pydub | MIT | The radio effect on Alan's voice |
| pywebview | BSD 3-Clause | The app's own windows |
| websockets | BSD 3-Clause | Live telemetry to the HUD |
| NumPy | BSD 3-Clause | Telemetry and audio maths |
| sounddevice | MIT | Microphone input |
| huggingface-hub · tokenizers | Apache 2.0 | Fetching the speech model |
| PyYAML | MIT | Reading the detection thresholds |
| python-dotenv | BSD 3-Clause | Reading your settings file |
| certifi | MPL 2.0 | Certificates for secure connections |
| Python | PSF | The language RadioCheck is written in |
Car and circuit data is derived in part from the community-maintained GT7 info database. Gran Turismo, PlayStation and PSVR2 are trademarks of their respective owners; RadioCheck is an independent project and is not affiliated with Sony Interactive Entertainment or Polyphony Digital.
© 2026 Rubén Martínez López · Privacy Policy · Terms of Use · Español